Evolution of the crystal stress distributions in face-centered cubic polycrystals subjected to cyclic loading

Evolution of the crystal stress distributions in face-centered cubic polycrystals subjected to cyclic loading

Available online at www.sciencedirect.com Acta Materialia 59 (2011) 6901–6916 www.elsevier.com/locate/actamat Evolution of the crystal stress distri...

2MB Sizes 0 Downloads 44 Views

Available online at www.sciencedirect.com

Acta Materialia 59 (2011) 6901–6916 www.elsevier.com/locate/actamat

Evolution of the crystal stress distributions in face-centered cubic polycrystals subjected to cyclic loading Su Leen Wong ⇑, Paul R. Dawson Sibley School of Mechanical and Aerospace Engineering, Cornell University, Ithaca, NY 14853, USA Received 30 April 2011; received in revised form 15 July 2011; accepted 17 July 2011

Abstract Due to the heterogeneous nature of polycrystalline metals, the stress distribution at the crystal level is influenced by the complex interplay of factors such as the orientations of the crystal lattices, the elastic and plastic mechanical properties, the interactions between neighboring crystals, and the type of loading conditions. In this paper, we investigate the evolution of the crystal scale stress and elastic strain distributions under cyclic loading. More specifically, we examine the orientation dependent lattice strains in face-centered cubic polycrystals under fully reversed, strain-controlled cyclic loading using a cyclic hardening model implemented within a crystal-based finite element formulation. The directional strength-to-stiffness ratio and the single crystal yield surface (SCYS) topology are used to provide a quantitative explanation of the observed hysteresis behavior when the lattice strains are plotted as a function of the macroscopic stress. The directional strength-to-stiffness ratio influences the progression of yielding in the elastic–plastic transition regime, leading to a nonlinear response in the elastic–plastic transition regime and causing hysteresis behavior in the macroscopic stress versus lattice strain curves. The lattice strains after the elastic–plastic transition regime are influenced by the single crystal elastic moduli and the facet/vertex attributes of the SCYS. We found that the same trends hold in the absence of slip system hardening and in the presence of a strong rolling texture. Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Finite element analysis; Elastic behaviour; Plastic deformation; Yield phenomena; Low-cycle fatigue

1. Introduction Recent advances in both diffraction methods and polycrystal deformation modeling have enabled researchers to better quantify stresses and strains at the crystal level. Using diffraction methods it is possible to measure elastic lattice strains within crystals under in situ loading. With parallel computing and data parallel implementations of structural codes, it is possible to construct and load polycrystalline aggregates with thousands of crystals, each crystal highly resolved by many finite elements. These advances offer the promise of being able to develop a quantitative link between macroscopic loading and spatial distribution ⇑ Corresponding author. Tel.: +1 607 280 8429.

E-mail addresses: [email protected] (S.L. Wong), [email protected] (P.R. Dawson).

of stress at the microstructural (crystal) scale under cyclic loading. Crystals within a polycrystalline aggregate tend not to share an applied load equally, but rather exhibit individual and heterogeneous responses that are dependent on the crystallographic orientation, the single crystal elastic and plastic anisotropy, and the interactions with neighboring crystals. It is generally accepted that these factors all play roles in determining crystal-scale stress distributions under cyclic loading, especially for the regime associated with low-cycle fatigue. However, a clear explanation is elusive due to the complexities of the modes and mechanisms of cyclic deformations. To this end, it is instructive to examine the heterogeneous nature of the crystal stresses with respect to the anisotropy inherent in the single crystal properties. A specific goal is to identify correlations between the single crystal properties and the orientation dependence of the crystal

1359-6454/$36.00 Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2011.07.042

6902

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

stresses in the presence of equilibrium and compatibility constraints associated with grain interactions. In particular, we examine two specific attributes of crystals: the directional strength-to-stiffness ratio and the single crystal yield surface (SCYS) topology. The directional strength-to-stiffness is an important metric for the relative load sharing among crystals of a polycrystal, especially in the elastic– plastic transition. The yield surface constrains the stress tensor, both in magnitude and direction, and as with the directional strength-to-stiffness its topology affects the nature of the distributions of stress among crystals. As will be presented here, both of these attributes have important roles with respect to cyclic loading behavior. The directional strength-to-stiffness provides a quantitative basis to explain the observed average lattice strains of crystals when loading proceeds into the elastic–plastic transition regime [1]. The orientation dependent lattice strain response observed in the elastic–plastic transition regime, typically referred to as intergranular strains, arise due to different sets of crystals yielding at different levels of the macroscopic stress. Load redistribution occurs between crystals of different orientations when some crystals begin to yield, forcing crystals that remain elastic to carry an increased proportion of the applied stress. The joint role of the elastic stiffness (elastic anisotropy) and yield strength (plastic anisotropy) on the development of these intergranular strains can be quantified using the directional strengthto-stiffness ratio. In general, crystals within a polycrystalline aggregate with lower directional strength-to-stiffness tend to yield before crystals with higher directional strength-to-stiffness. Here we demonstrate that the directional strength-to-stiffness ratio is a parameter that can be used to link the single crystal anisotropic properties and load sharing among differently orientated crystals acting collectively during cyclic loading. Over a wide range of temperatures and strain rates, crystals deform plastically by restricted slip on prescribed slip systems in prescribed directions [2]. The yield condition for a slip system can be described as a plane in fivedimensional (deviatoric) stress space and the rate independent SCYS is the inner envelope of the set of planes defined by all the slip systems of the single crystal [3]. A vertex of the SCYS is defined as the intersection of at least five of the slip system planes. For rate dependent behavior, the SCYS depends on strain rate and temperature. The rate dependent surfaces lie within the rate independent surface, but for materials with low rate dependence, the two surfaces are close. During deformation that includes slip, the SCYS is actively influencing the stress. As the stress level within a crystal rises, the number of slip systems activated increases, beginning with only one or two and eventually activating the number associated with a vertex of the SCYS, a condition referred to as polyslip [4]. The crystal stress direction changes as the number of constraints from activated slip systems increases. Once the stress aligns with a vertex of the SCYS and all potentially active slip systems have

been activated, plastic flow is fully developed. The influence of the vertices of the SCYS on the crystal stress distributions in fully developed plasticity has previously been discussed by Ritz et al. [5] for monotonic loading. Under cyclic loading, fully developed plasticity may not be reached before the loading reverses. Nevertheless, the topology of the SCYS influences the stress as soon as the stress exceeds the elastic limit and thus plays an important role in cyclic as well as monotonic loading behaviors. The directional strength-to-stiffness and the SCYS topology will be used to better understand stress distributions in polycrystals subjected to cyclic loading. Both are useful in quantifying the influence of crystal anisotropy on the orientation dependence of the stress distribution in the presence of grain interactions. The grain interactions arise due to requirements of equilibrium of forces and compatibility of motion, which are enforced in the modeling by means of a well-established finite element formulation. To demonstrate the concepts, we focus on providing a wellgrounded explanation of the orientation dependence of lattice strains observed in cyclic loading, especially their evolution over loading cycles. 2. Experimental observations Diffraction methods have the ability to interrogate the elastic lattice strains within crystals with orientations associated with a particular crystallographic fiber. Since the elastic strains in crystals are directly related to the stresses through Hooke’s law, diffraction experiments offer direct insight into the heterogeneous nature of the crystal stress distribution within a polycrystalline aggregate. Although many experimental studies measuring the evolution of lattice strains under monotonic tensile loading have been conducted on different types of polycrystalline materials [6–11], the number of studies involving lattice strain diffraction measurements of polycrystalline materials under fully reversed cyclic loading is very limited. Neutron diffraction has been used to measure the residual lattice strains in stainless steel under fully reversed, high cycle fatigue loading in load control [12,13]. High energy synchrotron X-rays have been used to track the evolution the lattice strains associated with a few crystallographic fibers under fully reversed cyclic loading in load control [14]. There have also been several studies based on using neutron diffraction techniques to measure the lattice strains for a limited number of crystallographic fibers under in situ, fully reversed, low cycle fatigue loading in strain control [15–18]. Examples of lattice strain data from fully reversed cyclic loading experiments that are relevant to our current numerical study are presented in Fig. 1. The lattice strain measurements shown in Fig. 1 were obtained from neutron diffraction experiments conducted by Lorentzen et al. [15] and Korsunsky et al. [16] on stainless steel under fully reversed cyclic loading. Lorentzen et al. [15] measured the

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

6903

Fig. 1. Lattice strains of crystals with the {1 1 1} and {2 0 0} lattice plane normals aligned with the axial loading direction for stainless steel specimens under fully reversed cyclic loading between fixed strain limits.

lattice strains for several crystallographic fibers aligned with the axial and transverse specimen directions in stainless steel under fully reversed cyclic loading. The experiment was conducted in incremental load control and the specimen was cycled at a fixed strain amplitude of ±0.4% for eight complete cycles. Diffraction measurements were conducted at 24 load increments throughout the cycle. Korsunsky et al. [16] used time-of-flight neutron diffraction to measure the lattice strains corresponding to two crystallographic fibers with the crystal directions aligned with the specimen axial loading direction for AISI 316L stainless steel. Two specimens were cycled at strain amplitudes of 0.25% and 0.44% in fully reversed strain-controlled cycling. Lattice strain measurements were conducted at eight macroscopic strain levels on each cycle. The measured lattice strains shown in Fig. 1, where the crystal directions are aligned with the specimen loading direction, exhibit an orientation dependent response over the course of a cycle. The lattice strains in crystals having {1 1 1} plane normals aligned with the specimen loading direction exhibit a close to linear response with increasing macroscopic stress. The lattice strains for crystals having {2 0 0} plane normals aligned with the specimen loading direction, on the other hand, exhibit significant hysteresis. The experimental data shown in Fig. 1 are an important example of how the lattice strains varying at the crystal level need to be explained in a consistent manner with respect to the single crystal anisotropic properties as part of any micromechanical description of fatigue processes. These experimental results motivate our investigation into the orientation dependent nature of load sharing during cyclic loading from the perspective of how the directional strength-to-stiffness and the SCYS topology influence the grain interactions and give rise to the behavior characterized by Fig. 1. We also examine the effect of texture and strain hardening on the progression of yielding through the elastic–plastic transition and on the evolution of the lattice strains during cyclic loading.

3. Methodology A crystal-based elastoplastic constitutive model implemented within a finite element framework is used to simulate the response of polycrystals under fully reversed straincontrolled cyclic loading. A complete description of the constitutive model and its implementation within a finite element framework can be found in Refs. [19–21]. The capabilities of the crystal elastoplastic constitutive model are only briefly summarized here:  multiplicative decomposition of the crystal deformation gradient into elastic, plastic and rotational portions;  anisotropic elasticity appropriate to the face-centered cubic (fcc) crystal structure;  plastic deformation by rate dependent restricted slip on the 12 {1 1 1}h1 1 0i slip systems associated with the fcc crystal structure;  strain hardening with pseudosaturation under cyclic loading;  texture evolution through crystallographic lattice reorientation. This crystal constitutive model is incorporated into a finite element formulation which has the following capabilities:  three-dimensional grain geometries, where each grain is resolved with multiple finite elements;  weak form of the equations of equilibrium;  implicit time-integration of the constitutive equations for numerical stability;  scalable parallel implementation using FORTRAN 90 and interprocessor communication based on the Message Passing Interface (MPI) standard. A cyclic slip system hardening model which tracks the accumulated slip system shear strains on individual slip

6904

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

systems and net crystal shearing rates within each crystal during a cycle [22] is used in the finite element model. Using a small modification to the conventional Voce hardening model, Turkmen et al. [22] were able to capture the pseudosaturation behavior observed in fully reversed cyclic loading experiments conducted on SS304L stainless steel. The evolution of the dislocation structure in the material is not explicitly modeled in the cyclic hardening model, but the net effect of the accumulation of dislocations on strain hardening is modeled implicitly through the evolution equations for the slip system strengths. The slip system and cyclic hardening parameters used in our simulations can be found in Ref. [22]. Simulations were conducted by deforming virtual polycrystals under fully reversed cyclic loading. Virtual polycrystals were instantiated with 1098 rhombic dodecahedra crystals (or grains), where each complete crystal was comprised of 48 10-node tetrahedra elements. There were also additional partial crystals which make up the surfaces of the finite element mesh, contributing to a total of 81,000 finite elements in the mesh. Mesh convergence studies were previously conducted by Miller et al. [23] using virtual specimens with 1098, 2916 and 10,976 complete dodecahedra crystals, which corresponds to 81,000, 192,000 and 648,000 elements, respectively. Only small differences in the crystal stress distributions were observed between the finite element meshes of varying sizes. We have also conducted a comparison of different mesh sizes in Ref. [1], using virtual specimens with 1098 and 2916 crystals. Quantities averaged along a crystallographic fiber were found to exhibit similar trends across different mesh sizes. On this basis, the mesh with 1098 complete crystals was used in the current study to permit a larger number of simulations to be performed. The simulation results can be compared to diffraction measurements of lattice strains by using crystallographic fibers as the basis of comparison. A crystallographic fiber, represented as cks, is the collection of all orientations with a particular crystal direction, c, aligned with a particular specimen direction, s. A crystallographic fiber consists of crystals with orientation, R, that satisfy the following equation: Rc ¼ s

ð1Þ

All orientations associated with a crystallographic fiber differ only by a rotation about a common axis. A crystallographic fiber corresponds directly to a particular Bragg diffraction condition, where c is the crystal {h k l} plane normal and s is a particular specimen direction. A cutoff angle of 5° between c and s is used in the current study to determine whether a crystal orientation belongs to a particular crystallographic fiber. The simulation data does not distinguish between lower-order and higher-order {h k l} reflections ({h k l} = n{h k l}). Therefore, only the lowestorder {hkl} lattice planes are considered. Details of the post-processing methodology has previously been reported in Ref. [1].

Two crystallographic fibers are examined in particular: the {1 0 0} k LD fiber and the {1 1 1} k LD fiber. The {1 0 0} k LD crystallographic fiber connects all crystals with orientations where the {1 0 0} crystal plane normal is aligned with the specimen loading direction (LD). The {1 0 0} k LD and {1 1 1} k LD fibers are chosen because they correspond to crystal orientations with extreme values of the directional strength-to-stiffness and these values vary significantly with rE. 4. Simulation results Three sets of simulations, designated as Sets 1, 2 and 3, were conducted using identical virtual polycrystals, except as noted below. For Set 1, the initial orientations of the crystals were randomly assigned from a uniform orientation distribution function (ODF). The cyclic slip system hardening model was active. For Set 2, the initial orientations of the crystals were randomly assigned from an ODF generated under plane strain compression to 70% strain. The ODF used in Set 2 is similar to the ODF shown in Ref. [24] and therefore will not be shown here. As with Set 1, the cyclic slip system hardening model was active for Set 2. For Set 3, the initial orientations of the crystals were the same as in Set 1. For this set, the cyclic slip system hardening model was inactive (no hardening). The different combinations of hardening and initial texture instantiated for each simulation set are shown in Table 1. Within each set of simulations, the only quantity that was varied was the single crystal elastic anisotropy ratio, rE, which is defined as: rE ¼

Eh111i Eh100i

ð2Þ

where Eh111i and Eh100i are the single crystal directional Young’s modulus in the h1 1 1i and h1 0 0i crystal directions, respectively. Two cases were considered: rE = 1.0 and rE = 3.2. For rE = 3.2, representative values of the single crystal elastic moduli (C11, C12 and C44) for stainless

Table 1 The different combinations of hardening and initial ODF for each set of simulations. Set

Hardening

Initial ODF

1 2 3

Yes Yes No

Uniform Plane strain compression Uniform

Table 2 Single crystal elastic moduli corresponding to rE = 1.0 and rE = 3.2. Representative values of the single crystal elastic moduli for stainless steel (rE = 3.2) were obtained from Ref. [25]. rE

C11 (GPa)

C12 (GPa)

C44 (GPa)

1.0 3.2

266.433 204.600

106.784 137.700

79.826 126.200

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

300

Macroscopic stress (MPa)

Macroscopic stress (MPa)

300

6905

200 100 0 −100

Cycle 0 Cycle 1 Cycles 2−20

−200 −300

200 100 0 r = 1.0

−100

E

r = 3.2 E

−200 −300

−1

−0.5

0

0.5

1

Macroscopic strain (%)

−1

−0.5

0

0.5

1

Macroscopic strain (%)

Fig. 2. Macroscopic true stress–strain curves for 20 complete cycles for simulation Sets 1 and 3.

steel were used [25]. For rE = 1.0, the moduli were adjusted to give isotropic single crystal behavior while leaving the macroscopic Young’s modulus the same as that for the rE = 3.2 case. The single crystal elastic moduli corresponding to these two values of rE are shown in Table 2. The method of varying the single crystal elastic moduli while keeping the average Young’s modulus of the aggregate constant is described in detail in Refs. [1,26]. The average shear modulus and Young’s modulus of the aggregate is not computed as an exact orientation average. However, the goal of the current study is to change the single crystal elastic moduli in such a way that the average Young’s modulus of the aggregate remains constant between the simulations within a set and not to use the simulations to attempt to match the actual average Young’s modulus of the material. While it is possible to use different methods for computing the average shear modulus and average Young’s modulus of the aggregate, the single crystal elastic moduli are changed to result in single crystal elastic anisotropy values of rE = 1.0 and rE = 3.2, regardless of the average shear modulus. Therefore the lattice strain behaviors will still exhibit the same orientation dependent response which can be predicted by the directional strength-to-stiffness ratio. Further, the computed uniaxial (tensile) loading stress–strain curves for the two rE values (and uniform textures) effectively overlap each other, which confirms that the method is giving us the intended outcome of varying the single crystal elastic moduli in a way that varies rE but not the average Young’s modulus of the aggregate. Fully reversed uniaxial tension–compression loading conditions were applied to the virtual polycrystals. A constant velocity was applied on the positive Z surface of the finite element mesh, while the Z = 0 face of the mesh was constrained in the Z-direction. The two positive X and Y surfaces of the mesh were traction free, while the X = 0 and Y = 0 surfaces had symmetry boundary conditions imposed. The simulations were conducted under fully reversed cyclic loading (R = 1) in strain control at a

constant strain rate of 0.001 s1 for 20 complete cycles. The virtual specimen was cycled between fixed macroscopic strain limits of ± 1% strain. 4.1. Macroscopic stress–strain response The macroscopic value of the axial stress was computed for the 20 loading cycles from the average normal traction over the loaded surface of the polycrystal. For Set 1, the resulting macroscopic stress for rE = 3.2 is plotted in Fig. 2a as a function of the macroscopic strain, which is computed on the basis of the overall axial dimension changes of the polycrystal. Cycles 0 and 1 are highlighted in Fig. 2a. Cycle 0 corresponds to monotonic loading in tension up to 1% macroscopic strain. The outermost loop corresponds to Cycle 20 in the deformation history. The loops grow in size with increasing cycles due to continued strain hardening. Since the polycrystals were cycled between fixed strain limits, a higher macroscopic stress level was required to achieve the same macroscopic strain limits on each cycle due to cyclic hardening. However, the cycle-by-cycle change in macroscopic stress at 1% macroscopic strain decreases with increasing cycles due to pseudosaturation of the slip system strength that is built into the model for the cyclic hardening behavior. With the exception of the elastic–plastic transition regime (the knee of the macroscopic stress–strain curve), the macroscopic stress–strain curves for the two values of rE in Set 1 are very similar. The macroscopic stresses for Set 2 are slightly higher overall than those of Set 1. This is because the plane strain compression texture used in the Set 2 simulations results in a higher average Taylor factor than the uniform ODF used for Set 1. Aside from the slight difference in the macroscopic stress level, the stress–strain hysteresis loops between Sets 1 and 2 are very similar and, consequently are not shown. Fig. 2b shows the macroscopic stress–strain curves for the two simulations of Set 3. In the absence of cyclic

6906

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

Fig. 3. Lattice strain, e{hkl}, vs. macroscopic true stress for the simulations of Set 1. The lattice strains are averaged among crystals belonging to the {1 0 0} k LD and {1 1 1} k LD crystallographic fibers, where LD corresponds to the [0 0 1] specimen loading direction.

300

400

{100} || LD {111} || LD

Macroscopic stress (MPa)

Macroscopic stress (MPa)

400

200 100 0 −100 −200 −300 −400 −3

−2

−1

0

1

2

3

300

{100} || LD {111} || LD

200 100 0 −100 −200 −300 −400 −3

−2

−1

0

1

2

3

Fig. 4. Lattice strain, e{hkl}, vs. macroscopic true stress for the simulations of Set 3. The lattice strains are averaged among crystals belonging to the {1 0 0} k LD and {1 1 1} k LD crystallographic fibers, where LD corresponds to the [0 0 1] specimen loading direction.

hardening, the slip system strength remains constant over the simulation and the macroscopic stress does not increase with cycles. Small differences between the macroscopic stress–strain curves are observed in the elastic–plastic transition as rE is varied. The slopes of the macroscopic stress– strain curves in the elastic loading and unloading portions of each cycle demonstrate that the single crystal elastic moduli were changed such that the macroscopic Young’s modulus is unchanged. 4.2. Lattice strain response The lattice strain, e{hkl}, for the {h k l} k LD crystallographic fiber is computed as follows: for each element in the mesh whose orientation belongs to the {h k l} k LD fiber to within a given tolerance, its full elastic strain tensor, , is

projected in the LD direction. The lattice strain value, {h k l}, is then obtained by averaging the projected elastic strain tensor value among all n crystals belonging to the {h k l} kLD fiber: ! n X T si  i  si =n efhklg ¼ ð3Þ i¼1

The average lattice strains associated with the {1 0 0} k LD and {1 1 1} k LD crystallographic fibers for Sets 1 and 3 are plotted in Figs. 3 and 4 with lattice strain on the abscissa and the macroscopic stress on the ordinate to facilitate comparison with the experimental data shown in Fig. 1. As with the macroscopic stress–strain hysteresis loops, there is little difference between the results for Sets 1 and 2 (uniform vs. plane strain compression textures), so only Set 1 is shown.

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

Examining first the lattice strains for Set 1, it is evident that the lattice strains for rE = 1.0, where the crystals are elastically isotropic, exhibit very different behaviors compared to the lattice strains for rE = 3.2, where the crystals have high elastic anisotropy. For rE = 1.0, shown in Fig. 3a, both the {1 0 0} k LD and {1 1 1} k LD lattice strains exhibit hysteresis over the course of a cycle. For rE = 3.2, shown in Fig. 3b, the {1 0 0} k LD lattice strains exhibit significant hysteresis, but the {1 1 1} k LD crystals exhibit a linear relation between the lattice strains and macroscopic stress over the course of a cycle. The lattice strain response without strain hardening (Set 3) is shown in Fig. 4. The lattice strain hysteresis loops do not increase in size with increasing cycles due to the absence of strain hardening. Nevertheless, the lattice strain hysteresis loops for Set 3 exhibit similar trends as the simulations in Sets 1 and 2. The difference in the single crystal elastic anisotropy can be observed from the slopes of the elastic loading and unloading portions of macroscopic stress vs. lattice strain curves. Note that for rE = 1.0, shown in Figs. 3a and 4a, the {1 0 0} k LD and {1 1 1} k LD lattice strains have the same slope in the elastic loading and unloading portions of the cycle, which is consistent with the isotropic nature of the crystals. For rE = 3.2 however, where the lattice strains are shown in Figs. 3b and 4b, the lattice strains in the elastic loading and unloading portion of the cycle exhibit different slopes, illustrating the influence of elastic anisotropy on the {1 0 0} k LD and {1 1 1} k LD crystals. We can compare the results for rE = 3.2 with the experimental data. The computed lattice strains shown in Fig. 3b are qualitatively consistent with the measured lattice strains shown in Fig. 1. The near-linear response of the {1 1 1} k LD lattice strains with macroscopic stress, with no evidence of hysteresis loops, observed in the measurements is predicted in the simulations. Further, the computed {2 0 0} k LD lattice strains exhibit significant hysteresis, which also agrees qualitatively with the measured {2 0 0} k LD lattice strains. However, the computed {2 0 0}k LD lattice strains exhibit a gradual downward inflection in the elastic–plastic transition regime which is not evident in the measured {2 0 0}k LD lattice strains. This discrepancy between the computed and measured {2 0 0}k LD lattice strains has also been observed by Lorentzen et al. [15] when using a selfconsistent elastoplastic model. Lorentzen et al. [15] attributed this to the lack of experimental data in the region just past zero macroscopic stress, since diffraction measurements were conducted at only 24 points during a cycle. The lattice strains measurements by Korsunsky et al. [16] shown in Fig. 1b were conducted at eight points during a cycle. Given the rapid changes occurring during the elastic–plastic transition regime, the frequency of measurements is not sufficient to capture details of the elastic– plastic transition and we do not expect a close match between the measured and computed lattice strains in this regime.

6907

4.3. Slip system activity Although the {1 1 1} k LD lattice strains for rE = 3.2 exhibit linear behavior between the elastic lattice strain and the macroscopic stress over the course of a cycle, this is not indicative of these crystals deforming within their elastic regime. To demonstrate this, we examine the plastic deformation rate magnitude, Dpfhklg , cycle by cycle for the crystals belonging to both the {1 0 0} k LD and {1 1 1} k b 0p , is a LD fibers. The plastic deformation rate vector, D five-dimensional deviatoric vector computed at the centroid of each element in the mesh. The second-order tensor b 0p for a single element in the mesh is computed as: form of D X bp ¼ ^ aÞ ð4Þ D c_ a symð^sa  m a

where c_ a is the slip system shearing rate, ^sa is the slip direc^ a is the slip plane normal for each a-slip system. tion, and m b 0p is computed The vector norm for the deviatoric vector D and averaged among all crystals belonging to the {h k l} k LD fiber to obtain the plastic deformation rate magnitude, Dpfhklg . For each cycle, we track Dpfhklg starting at a macroscopic stress of 100 MPa and continuing until a macroscopic strain of 1% is achieved, as indicated in Fig. 5. The results for Sets 1 and 3 are presented in Figs. 6 and 7, respectively. Again, Set 2 is similar to Set 1 and is not shown. A steep increase in Dpfhklg at a particular macroscopic stress indicates that on average, the crystals associated with the {h k l} k LD fiber have yielded and plastic flow is occurring among those crystals. The results indicate that the crystals belonging to both the {1 0 0}k LD and {1 1 1} k LD fibers experience plastic deformation over each cycle, even though there is no hysteresis associated with the {1 1 1} k LD lattice strains, as shown in Figs. 3b and 4b. Only a constant Dpfhklg value of zero over the course of a cycle for all macroscopic stress values would indicate that the crystals are remaining fully elastic.

Fig. 5. The solid black lines on the stress–strain curve correspond to the portions of the stress–strain curve where Dpfhklg is plotted for Cycles 0, 1 and 20 for Set 1.

6908

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

2.0

× 10

−3

2.0

{100} || LD, Cycle 0 {111} || LD, Cycle 0 {100} || LD, Cycles 1−20 {111} || LD, Cycles 1−20

1.8 1.6

−1 (s )

1.4

{hkl}

−1

{100} || LD, Cycle 0 {111} || LD, Cycle 0 {100} || LD, Cycles 1−20 {111} || LD, Cycles 1−20

1.6

Dp

(s )

p {hkl}

D

1.0

−3

1.8

1.4 1.2

× 10

0.8

1.2 1.0 0.8

0.6

0.6

0.4

0.4

0.2

0.2

0 −100

0

100

200

0 −100

300

0

Macroscopic stress (MPa)

100

200

300

Macroscopic stress (MPa)

Fig. 6. Plastic deformation rate magnitude, Dpfhklg , vs. macroscopic true stress for crystals belonging to the {1 0 0} k LD and {1 1 1} k LD crystallographic fibers for the simulations of Set 1. The LD direction corresponds to the [0 0 1] specimen loading direction. The plastic deformation rate magnitude, Dpfhklg , is plotted starting at a macroscopic stress of 100 MPa until a macroscopic strain of 1% is achieved, which is shown for cycles 0, 1 and 20 in Fig. 5.

2.0

× 10

1.8 1.6

−3

2.0

{100} || LD, Cycle 0 {111} || LD, Cycle 0 {100} || LD, Cycles 1−20 {111} || LD, Cycles 1−20

1.8 1.6

−1

1.2 1.0

D

(s )

−1

0.8

{100} || LD, Cycle 0 {111} || LD, Cycle 0 {100} || LD, Cycles 1−20 {111} || LD, Cycles 1−20

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0 −100

−3

1.4

p {hkl}

(s )

1.0

D

p {hkl}

1.4 1.2

× 10

0

100

200

300

Macroscopic stress (MPa)

0 −100

0

100

200

300

Macroscopic stress (MPa)

Fig. 7. Plastic deformation rate magnitude, Dpfhklg , vs. macroscopic true stress for crystals belonging to the {1 0 0} k LD and {1 1 1} k LD crystallographic fibers for the simulations of Set 3. The LD direction corresponds to the [0 0 1] specimen loading direction.

Figs. 6a and 7a demonstrate that for rE = 1.0, even though the crystals are elastically isotropic, the {1 0 0} k LD and {1 1 1} k LD crystals yield on average at different macroscopic stress levels. The difference in the macroscopic stress at which yielding occurs on average in these crystals can therefore be attributed to the plastic anisotropy (orientation dependent yield strength). Since the {1 0 0} k LD crystals yield on average at a lower macroscopic stress compared to the {1 1 1} k LD crystals, it implies that for rE = 1.0, the {1 0 0} k LD crystals on average are more favorably oriented for yielding than the {1 1 1} k LD crystals. When rE is increased to rE = 3.2, the {1 0 0} k LD crystals now yield at a higher macroscopic stress on average compared to the {1 1 1} k LD crystals. Therefore, the value

of rE influences the macroscopic stress at which yielding begins for crystals associated with different fibers. This implies that the progression of yielding among these crystals is influenced by both the elastic and plastic properties, which will be examined in more detail in Section 5.2. It is also observed from Figs. 6 and 7 that the macroscopic stress at which yielding begins is lower on Cycle 1 compared to Cycle 0. This phenomena arises due to yield asymmetry, which is a Bauschinger-type phenomena first observed by Czyzak et al. [27] and Hutchinson et al. [28]. Barton et al. [29], using an elastoplastic crystal-based finite element model similar to the one used in our current study, demonstrated that a polycrystal yields at a lower macroscopic stress in compression after it has been yielded in

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

tension due to residual crystal stresses that are present even in the macroscopically unloaded state. After Cycle 1, however, the macroscopic stress at which yielding begins increases with increasing cycles due to cyclic hardening. Yield asymmetry can still be observed even when the cyclic slip hardening model is inactive for Set 3, as seen in Fig. 7, although the macroscopic stress at which yielding begins does not increase with cycles after Cycle 1. 5. Influences of directional strength-to-stiffness and SCYS topology on the cyclic response 5.1. Structural analog The concept of the strength-to-stiffness ratio is first demonstrated using a simple isostrain structural analog in Fig. 8. Fig. 8a shows two blocks of different materials but with identical dimensions loaded in compression between two rigid plates, deforming with identical strains. The two different materials have different Young’s modulus (E1 and E2), and different yield strengths (Y1 and Y2) which result in different values of their strength-to-stiffness ratios, Y1/E1 and Y2/E2. The two materials are assumed to have elastic–perfectly plastic behavior. Since Material 1 has a lower yield strength compared to Material 2, we would expect it to yield before Material 2 as they are loaded. However, Material 2 yields first because by having a higher Young’s modulus it reaches its yield strength before Material 1, as shown in Fig. 8b, which is a schematic showing the individual stress–strain response of each material. This behavior indicates that the Young’s modulus, in addition to the yield strength, determines the relative order the two materials will yield. Material 2, in fact, has the lower strength-to-stiffness (Y/E) ratio and it yields before Material 1. Material 1 has the higher strength-to-stiffness ratio even though it has lower yield strength. Therefore, high Y/E ratio leads to the material yielding relatively later in a system having elements whose values of Y/E differ.

(a)

6909

5.2. Directional strength-to-stiffness To explain why the lattice strain hysteresis loops differ qualitatively with rE, we consider the orientation dependence of both the elastic and plastic properties of the crystals. Although there exist minor differences stemming from the initial texture and from the absence or presence of hardening, a consistent trend emerged from the three sets of simulations. The {1 0 0} k LD crystals yield on average at a lower macroscopic stress compared to the {1 1 1} k LD crystals when rE = 1.0, as shown in Figs. 6a and 7a. However, for rE = 3.2, the reverse situation occurs, where the {1 1 1} k LD crystals yield on average at a lower macroscopic stress compared to the {1 0 0} k LD crystals, which is evident from Figs. 6b and 7b. Thus, the macroscopic stress at which yielding begins for anisotropic crystals is influenced by the combination of the elastic and plastic anisotropy of the crystals, which can be quantified using the directional strength-to-stiffness ratio [1]. The directional strength-to-stiffness ratio for a single crystal under uniaxial tension applied in the direction, d, relative to the crystal lattice, rcrys S , is defined as: rcrys ¼ S

1 md  Ed

ð5Þ

where md is the Schmid factor [30] and Ed is the single crystal directional Young’s modulus [31] for the crystal direction d. The Schmid factor, md, quantifies the relative ease for single slip to occur in a single crystal with a uniaxial stress applied in the crystal direction d. Since a high value of md indicates that a particular crystal direction is more favorably oriented for slip, the directional strength is defined as the inverse of md in Eq. (5). The directional stiffness is defined as Ed in the direction d. The orientation dependence of the single crystal directional strength-to-stiffness, rcrys S , is shown within the basic cubic orientation triangle in Fig. 9. We have previously demonstrated that for a polycrystalline aggregate under uniaxial tension, crystals with high

(b)

Fig. 8. (a) Two blocks of identical dimensions but of different materials are loaded in compression between rigid plates which result in an isostrain condition for the two materials. (b) Individual stress–strain response of each material. Material 2 achieves its yield strength before Material 1 under the same strain even though Material 2 has the higher yield strength.

6910

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916 〈111〉

〈100〉

〈111〉

0.026

〈110〉

0.026

0.024

0.024

0.022

0.022

0.020

0.020

0.018

0.018

0.016

0.016

0.014

0.014

0.012

0.012

0.010

〈100〉

〈110〉

0.010

Fig. 9. Single crystal directional strength-to-stiffness ratio shown within the basic orientation triangle for rE = 1.0 and rE = 3.2. The figures have the same grayscale bar for comparison purposes.

directional strength-to-stiffness ratio will yield on average at a relatively higher macroscopic stress compared to crystals with low directional strength-to-stiffness ratio [1]. Fig. 9a shows that for rE = 1.0, the h1 1 1i direction has the higher strength-to-stiffness ratio relative to the h1 0 0i direction. Conversely, Fig. 9b shows that for rE = 3.2, the h1 0 0i direction instead has the higher strength-to-stiffness relative to the h1 1 1i direction. As rE is increased, the h1 0 0i strength-to-stiffness increases while the h1 1 1i strength-to-stiffness decreases. The same trend exists for average strength-to-stiffness ratios taken over all the orientations lying along a crystallographic fiber, as reported in Ref. [1]. Table 3 gives the strength-to-stiffness values for the {1 0 0} k LD and {1 1 1} k LD fibers along with the corresponding single crystal values. Comparing Figs. 6–9, the macroscopic stress at which the {1 0 0} k LD crystals begin yielding increases as the h1 0 0i strength-to-stiffness increases. Conversely, the macroscopic stress at which the {1 1 1} k LD crystals begin yielding decreases as the h1 1 1i strength-to-stiffness decreases. Therefore, the macroscopic stress at which yielding begins on average for crystals associated with a crystallographic fiber correlates well with the directional strengthto-stiffness ratio at low and high values of rE. For crystals with intermediate values of rE, particularly in the range of rE = 1.5  2.0, the h1 0 0i and h1 1 1i crystals have similar directional strength-to-stiffness values [1]. At intermediate values of rE, the interaction with neighboring crystals together with the directional strength-to-stiffness ratio are both important influences on the stress distribution of the crystals associated with a fiber.

We now focus on the results from the simulations of Set 1 to examine the role of the directional strength-to-stiffness in determining the lattice strain and plastic deformation rate response for crystals associated with the {1 0 0} k LD and {1 1 1} k LD fibers. For both values of rE, the crystals that yield on average at a relatively higher macroscopic stress (Fig. 6) also exhibit a downward inflection in lattice strain (Fig. 3). For rE = 1.0, the {1 1 1} k LD crystals yield at a higher macroscopic stress which is accompanied by a downward inflection in the lattice strain. For rE = 3.2, the {1 0 0} k LD crystals instead yield at a higher macroscopic stress which is also accompanied by a downward inflection in the lattice strain. The downward inflection of the lattice strains in the elastic–plastic transition regime is due to redistribution of additional load between crystals that have yielded and crystals that have not yielded [1,32], where proportionally more elastic straining can occur in crystals that have not yielded. As rE is increased from the isotropic case (rE = 1.0) to the highly anisotropic case (rE = 3.2), the behavior of the {1 1 1} k LD lattice strains changes from a downward inflection to near-linear behavior with increasing macroscopic stress. Conversely, the {1 0 0} k LD lattice strains exhibit an upward inflection for rE = 1.0 which becomes a downward inflection for rE = 3.2. This is a consequence of the change in the h1 0 0i directional strength-to-stiffness from low to high relative to that of other orientations which changes the macroscopic stress at which yielding begins for the {1 0 0} k LD crystals.

Table 3 Directional strength-to-stiffness values for selected single crystal directions and averages over crystallographic fibers.

Plastic deformation via slip initiates when the crystal stress reaches the SCYS. With increasing plastic strain, the crystal stress traverses the SCYS until it reaches a vertex [3,29]. The crystal stress moves toward a vertex to accommodate a deviatoric strain state that is compatible with the motion of the polycrystal. The SCYS defined by

rE

h1 0 0i

h1 1 1i

{1 0 0} k LD

{1 1 1} k LD

1.0 3.2

0.0119 0.0261

0.0179 0.0123

0.0113 0.0213

0.0236 0.0177

5.3. Vertices of the single crystal yield surface

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916 Table 4 Relative magnitudes of the deviatoric vertex stress, r0v , for the five vertex families associated with the fcc SCYS [5]. Vertex family number

Relative magnitude of r0v

V1 V2 V3 V4 V5

1.00 1.50 1.73 1.33 1.23

the 12 {1 1 1}h1 1 0i slip systems in fcc crystals has 56 vertices (28 positive vertices). The vertices can be grouped into five unique vertex families due to crystal symmetries, where the vertex stresses within a vertex family are indistinguishable from each other [33]. The relative magnitudes of the

vertex stresses for the five families are listed in Table 4. An important aspect of the behavior of fcc polycrystals is that under monotonic tensile loading, the crystal stresses associated with a particular crystallographic fiber tend to align more closely with certain vertex families than others in fully developed plasticity [5]. Here, we examine the lattice strains with respect to the five unique vertex families of the SCYS for fully reversed cyclic loading. For Set 1, the {1 0 0} k LD and {1 1 1} k LD lattice strains are plotted as a function of the macroscopic strain for rE = 1.0 and rE = 3.2 in Fig. 10. First, note that the initial slopes of the curves during the elastic loading and unloading portions of the straining cycle reflect the elastic anisotropy. For rE = 1.0, the {1 0 0} k LD and {1 1 1} k LD lattice strains have the same slopes as each other during

−3

3

−3

x 10

3

{100} || LD {111} || LD

x 10

{100} || LD {111} || LD

{hkl}

2

1

Lattice strain, ε

Lattice strain, ε

{hkl}

2

0 −1 −2 −3

6911

1 0 −1 −2

−1

−0.5

0

0.5

Macroscopic strain (%)

1

−3

−1

−0.5

0

0.5

1

Macroscopic strain (%)

Fig. 10. Lattice strain, e{hkl}, vs. macroscopic true strain associated with the {1 0 0} k LD and {1 1 1} k LD fibers for the simulations of Set 1, where LD is the specimen loading direction.

Fig. 11. Coaxiality angle, /vc , between the crystal stress and the nearest vertex stress for several crystallographic fibers at 1% macroscopic strain on Cycle 1. V1–V5 correspond to the five unique vertex families.

6912

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

Fig. 12. Points along the macroscopic stress–strain curve on Cycles 0, 1 and 20 for which the distributions of the normalized deviatoric stress magnitudes are plotted in Figs. 13–15.

both the elastic loading and unloading portions of the cycle. For rE = 3.2, however, the lattice strains for these two fibers exhibit different slopes at the start of the elastic loading and unloading portions of each cycle, as expected from the single crystal elastic moduli. After the elastic– plastic transition regime, but before unloading, the straining of the sample is accommodated predominantly by plastic straining and the lattice strains approach approximately

steady values consistent with fully developed plasticity. (The lattice strains do change slightly between the elastic– plastic transition and unloading, but the changes are small in comparison to the changes in the elastic regime.) For rE = 1.0, shown in Fig. 10a, the steady value of the {1 1 1} k LD lattice strains is larger than the steady value of the {1 0 0} k LD lattice strains. In contrast, for rE = 3.2 the reverse is true, as is evident from Fig. 10b. For the elastically isotropic case of rE = 1.0, the relative differences in the lattice strains in fully developed plasticity are a consequence of the SCYS. The crystal stresses are on average different for the two fibers. For the elastically anisotropic case of rE = 3.2, the relative differences in lattice strains are a consequence of both the SCYS and the elastic anisotropy. The average moduli, as well as the crystal stresses, are on average different for the two fibers. Using an analysis similar to that conducted by Ritz et al. [5], we can show that the average stresses for the two fibers reflect the constraint imposed by the SCYS. The angle of coaxiality, /vc , between the deviatoric form of the crystal stress, r0c , and the deviatoric form of the nearest vertex stress, r0v , is computed for each finite element associated with a crystallographic fiber as follows: /vc ¼ cos1



r0c  r0v kr0c kkr0v k



Fig. 13. Normalized deviatoric stress magnitudes for rE = 1.0 and rE = 3.2 on Cycle 0 at 0.065%, 0.4% and 1% macroscopic strain.

ð6Þ

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

6913

Fig. 14. Normalized deviatoric stress magnitudes for rE = 1.0 and rE = 3.2 on Cycle 1 at 0.8%, 0.3% and 1% macroscopic strain.

A coaxiality angle of zero indicates that the crystal stress is completely aligned with its preferred vertex stress. The /vc distributions corresponding to several crystallographic fibers are shown at 1% macroscopic strain on Cycle 1 for rE = 1.0 and rE = 3.2 in Figs. 11a and 11b, respectively. The distributions are very insensitive to the elastic anisotropy. The /vc distributions for Cycle 20 are very similar to the /vc distributions for Cycle 1, even with the stress increasing on successive cycles due to strain hardening. For both Cycle 1 and 20, the crystal stresses for the various fibers associate strongly with particular vertex families. The slip system strengths within a crystal evolve isotropically, resulting in an increase in the size of the SCYS without a change in its shape. The preferred vertices for each fiber thus remain unchanged with increasing numbers of cycles. This observation for fully reversed cyclic loading is consistent with previous results discussed by Ritz et al. for fcc polycrystals under monotonic tensile loading [5]. We now return to the lattice strains presented in Fig. 10. At 1% macroscopic strain on a particular cycle, the crystal stress distributions over the polycrystal are approximately the same for rE = 1.0 and rE = 3.2, but the magnitudes of lattice strains differ. This is expected from the differences in the elastic moduli for the two levels of elastic anisotropy. For rE = 1.0, the difference between the {1 0 0} k LD and {1 1 1} k LD lattice strains in Fig. 10a is due solely to the

difference between the magnitudes of the V1 and V2 vertex stresses. From Fig. 11, the preferred vertex family of the {1 0 0} k LD crystals is the V1 vertex family and the preferred vertex family of the {1 1 1} k LD crystals is the V2 vertex family. Consequently, for rE = 1.0, the {1 1 1} k LD steady lattice strains are larger than the {1 0 0} k LD steady lattice strains since the magnitude of the V2 vertex stress is larger than the V1 vertex stress, which is demonstrated in Table 4. The difference in the lattice strains for rE = 3.2, however, which are shown in Fig. 10b, is due to a combination of the single crystal elastic moduli and the difference between the V1 and V2 vertex stresses. The coaxiality angle distributions indicate that the crystal stresses near the end of each loading cycle tend to align with the same sets of SCYS vertices irrespective of the elastic anisotropy and the cycle number. This trend holds true for simulation Sets 1, 2 and 3. However, variability does exist in the stress magnitude among the crystals associated with a particular fiber. This variability is a function both of the elastic anisotropy and the cycle number. To examine the stress variability, crystal stress frequency distributions are constructed for the crystals associated with a particular fiber. Specifically, for all the finite elements with lattice orientation lying along a designated fiber, we construct the frequency distribution of the magnitude of the crystal deviatoric stress vector normalized with respect to the

6914

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

Fig. 15. Normalized deviatoric stress magnitudes for rE = 1.0 and rE = 3.2 on Cycle 20 at 0.8%, 0.3% and 1% macroscopic strain.

magnitude of the macroscopic deviatoric stress. The evolution of these frequency distributions over the course of the loading cycle, as indicated on Fig. 12, is examined for Cycles 0, 1 and 20, for both values of rE and for several crystallographic fibers. The normalized stress distributions on Cycle 0 for rE = 1.0 and rE = 3.2 are shown in Figs. 13a and 13b, respectively, at the three points along the macroscopic stress–strain curve. These correspond to 0.065%, 0.4% and 1% macroscopic strain. In the elastic regime at 0.065% macroscopic strain, the normalized stress distributions differ significantly between rE = 1.0 and rE = 3.2, reflecting the difference in elastic anisotropy of the crystals. As the deformation proceeds, the crystal stresses evolve through the elastic–plastic transition toward similar distributions at 1% macroscopic strain. It is observed from the macroscopic stress–strain curve in Fig. 12 that at 1% macroscopic strain for Cycle 0, the material is continuing to strain harden, which may cause differences in the stress distributions between rE = 1.0 and rE = 3.2 because the crystals may not be experiencing fully developed plastic flow at this stage of the deformation. The normalized stress distributions for Cycle 1 are plotted in Fig. 14 at 0.8%, 0.3% and 1% macroscopic strain, as indicated in Fig. 12. On this cycle, as well as all succeeding cycles, the normalized stress distributions at 0.8% macroscopic strain (prior to the onset of the elastic–plastic

transition) are not homogeneous, even for isotropic elasticity (rE = 1.0). This is due to the presence of residual stresses induced by plastic straining on Cycle 0. The distributions at this stage of the deformation differ between rE = 1.0 and rE = 3.2 due to the influence of the directional strengthto-stiffness on load sharing among crystals, as discussed previously in Section 5.2. The presence of residual stresses as well as the directional strength-to-stiffness ratio influence the crystal stresses in the elastic–plastic transition as the deformation proceeds. With increasing macroscopic strain, the normalized stress distributions for both rE = 1.0 and rE = 3.2 evolve toward a similar pattern in fully developed plasticity, as demonstrated by comparing Figs. 14a and 14b at 1% macroscopic strain. At this point the stress distributions are dominated by the SCYS, which are the same for the two simulations. The normalized stress distributions for Cycle 20, shown in Fig. 15, exhibit some similarities and differences to the normalized stress distributions for Cycle 1. First, note that the normalized stress distributions for Cycles 1 and 20 at the point just prior to the knee of the macroscopic stress– strain curve (0.8% macroscopic strain) are not directly comparable. Even though these two points on different cycles are at the same macroscopic strain level, they may correspond to different stages in the deformation. The macroscopic stresses differ between these two points in part due to continued cyclic hardening. The normalized stress

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

6915

Fig. 16. Slip system strength distribution by crystallographic fiber at 1% macroscopic strain for rE = 1.0 and rE = 3.2.

distributions for Cycles 1 and 20 show similar trends in that the level of the elastic anisotropy influences the frequency distributions at 0.8% and 0.3% macroscopic strain, but not at 1.0% macroscopic strain. The normalized stress distributions at 1.0% macroscopic strain on Cycles 1 and 20 differ from each other in terms of the spread of the stresses. With increased cycles, the standard deviations of the stress distributions increases for all fibers, which arises due to the evolution of the slip system strengths on each cycle. Because there exists variation in the crystal stresses, there also exists variation in the slip system shearing rates. Thus, the standard deviations of the slip system strengths also increase, as shown in Fig. 16. Over the course of the cyclic loading, this leads to greater variation in the normalized stresses at 1% macroscopic strain, which is shown in Figs. 13–15.





6. Conclusions Several conclusions can be obtained from the simulations conducted in the current study. These are summarized as follows:  Lattice strain hysteresis behavior during cyclic loading exhibits an orientation dependent as well as an elastic anisotropy-dependent response. Hysteresis loops are evident in lattice strains vs. macroscopic stress records.





The loops differ in slope and the degree to which they are open or closed. Both depend on the crystallographic fiber. The lattice strains for a particular fiber may exhibit a linear relation with macroscopic stress, which is characterized by a closed hysteresis loop. This behavior does not necessarily indicate that the crystals remain elastic over the whole cycle, as is demonstrated by examining the corresponding plastic deformation rates averaged along a fiber. The directional strength-to-stiffness ratio can be used to explain the lattice strain hysteresis behavior as well as the progression of yielding through the elastic–plastic transition for different fibers. In general, high directional strength-to-stiffness corresponds to the crystals yielding at a higher macroscopic stress, causing the crystals which have not yielded to exhibit an upward inflection in the lattice strains in the elastic–plastic transition regime since they are able to carry a higher proportion of the applied load. Open hysteresis loops indicate a difference in relative strength-to-stiffness from the average. Closed hysteresis loops indicate a relative strength-to-stiffness that is close to the average. After the elastic–plastic transition regime, crystals belonging to certain fibers tend to align more closely with certain vertex families of the SCYS. These crystal

6916

S.L. Wong, P.R. Dawson / Acta Materialia 59 (2011) 6901–6916

stresses in fully developed plasticity are not dependent on the elastic anisotropy. The lattice strains in this regime, however, are influenced by the elastic anisotropy and the relative magnitudes of the SCYS vertex stresses.  The distribution of crystal stress magnitudes in the elastic and elastic–plastic transition regime are influenced by the elastic anisotropy and the residual stresses. Even with continued strain hardening with cycles, these crystal stresses on each cycle evolve toward the same SCYS vertex families in fully developed plasticity irrespective of the elastic anisotropy.  The standard deviation of the crystal stress magnitudes for a fiber increases with increasing cycles due to the different boundary conditions experienced by the crystals on each cycle with continued cyclic loading, leading to increased variation with cycles in the slip system strengths for the crystals belonging to a fiber.  These results were found not to be significantly influenced by the absence of slip system hardening and in the presence of a strong rolling texture. However, these results hold only for an isotropic slip system hardening model because an anisotropic slip system hardening model may influence the evolution of the slip system strengths with cycles.

Acknowledgement Support for this work has been provided by the Air Force Office of Scientific Research (AFOSR) under Grant No. FA9550-06-1-0168. References [1] Wong SL, Dawson PR. Influence of directional strength-to-stiffness on the elastic–plastic transition of fcc polycrystals under uniaxial tensile loading. Acta Mater 2010;58(5):1658–78. [2] Frost HJ, Ashby MF. Deformation-mechanism maps: the plasticity and creep of metals and ceramics. Pergamon; 1982. [3] Kocks UF. The relation between polycrystal deformation and singlecrystal deformation. Metall Trans 1970;1:1121–43. [4] Kocks UF. Polyslip in single crystals. Acta Metall 1960;8(6):345–52. [5] Ritz H, Dawson PR, Marin T. Analyzing the orientation dependence of stresses in polycrystals using vertices of the single crystal yield surface and crystallographic fibers of orientation space. J Mech Phys Solids 2010;58(1):54–72. [6] Holden TM, Clarke AP, Holt RA. Neutron diffraction measurements of intergranular strains in MONEL-400. Metall Mater Trans A 1997;28A:2565–76. [7] Pang JWL, Holden TM, Mason TE. In situ generation of intergranular strains in an Al7050 alloy. Acta Mater 1998;46(5):1503–18. [8] Clausen B, Lorentzen T, Bourke MAM, Daymond MR. Lattice strain evolution during uniaxial tensile loading of stainless steel. Mater Sci Eng A 1999;259(1):17–24. [9] Clausen B, Bourke M. Lattice plane response during tensile loading of an aluminum 2 percent magnesium alloy. Metall Mater Trans A 2001;32(13):691–4. [10] Daymond MR, Tome´ CN, Bourke MAM. Measured and predicted intergranular strains in textured austenitic steel. Acta Mater 2000;48(2):553–64.

[11] Brown DW, Bourke MAM, Clausen B, Holden TM, Tome´ CN, Varma R. A neutron diffraction and modeling study of uniaxial deformation in polycrystalline beryllium. Metall Mater Trans A 2003;34A:1439–49. [12] Wang YD, Tian H, Stoica AD, Wang X-L, Liaw PK, Richardson JW. The development of grain-orientation-dependent residual stresses in a cyclically deformed alloy. Nat Mater 2003;2:101–6. [13] Wang X-L, Wang YD, Stoica AD, Horton DJ, Tian H, Liaw PK, et al. Inter- and intragranular stresses in cyclically-deformed 316 stainless steel. Mater Sci Eng A 2005;399:114–9. [14] Park J-S, Revesz P, Kazimirov A, Miller MP. A methodology for measuring in situ lattice strain of bulk polycrystalline material under cyclic load. Rev Sci Instrum 2007;78(2):023910. [15] Lorentzen T, Daymond MR, Clausen B, Tome´ CN. Lattice strain evolution during cyclic tensile loading of stainless steel. Acta Mater 2002;50:1627–38. [16] Korsunsky AM, James KE, Daymond MR. Intergranular stresses in polycrystalline fatigue: diffraction measurement and self-consistent modelling. Eng Fract Mech 2004;71:805–12. [17] Huang E-W, Clausen B, Wang YD, Choo H, Liaw PK, Benson ML, et al. A neutron-diffraction study of the low-cycle fatigue behavior of HASTELLOYÒ C-22HSe alloy. Int J Fatigue 2007;29(9–11):1812–9. [18] Huang E-W, Barabash RI, Wang YD, Clausen B, Li L, Liaw PK, et al. Plastic behavior of a nickel-based alloy under monotonictension and low-cycle-fatigue loading. Int J Plast 2008;24(8):1440–56. [19] Marin EB, Dawson PR. On modelling the elasto-viscoplastic response of metals using polycrystal plasticity. Comput Methods Appl Mech Eng 1998;165(1–4):1–21. [20] Marin EB, Dawson PR. Elastoplastic finite element analyses of metal deformations using polycrystal constitutive models. Comput Methods Appl Mech Eng 1998;165(1–4):23–41. [21] Han T-S, Dawson PR. A two-scale deformation model for polycrystalline solids using a strongly-coupled finite element methodology. Comput Methods Appl Mech Eng 2007;196(13–16):2029–43. [22] Turkmen HS, Miller MP, Dawson PR, Moosbrugger JC. A slip-based model for strength evolution during cyclic loading. J Eng Mater Technol 2004;126:329–38. [23] Miller MP, Park J-S, Dawson PR, Han T-S. Measuring and modeling distributions of stress state in deforming polycrystals. Acta Mater 2008;56:3927–39. [24] Mika DP, Dawson PR. Effects of grain interaction on deformation in polycrystals. Mater Sci Eng A 1998;257(1):62–76. [25] Ledbetter HM. In: Levy M, Bass HE, Stern RR, editors. Handbook of elastic properties of solids, liquids, gases, vol. III. Academic; 2000. p. 291–8. 313–24. [26] Dawson PR, Boyce DE, MacEwen SR, Rogge RB. On the influence of crystal elastic moduli on computed lattice strains in AA-5182 following plastic straining. Mater Sci Eng A 2001;313(1–2):123–44. [27] Czyzak SJ, Bow N, Payne H. On the tensile stress–strain relation and the Bauschinger effect for polycrystalline materials from Taylor’s model. J Mech Phys Solids 1961;9(1):63–6. [28] Hutchinson JW. Plastic stress–strain relations of fcc polycrystalline metals hardening according to Taylor’s rule. J Mech Phys Solids 1964;12(1):11–24. [29] Barton NR, Dawson PR, Miller MP. Yield strength asymmetry predictions from polycrystal elastoplasticity. Transactions of the ASME 1999;121:230–9. [30] Schmid E, Boas W. Plasticity of crystals with special reference to metals. CRC Press; 1968. [31] Hosford WF. The mechanics of crystals and textured polycrystals. Oxford Science Publications; 1993. [32] Clausen B, Lorentzen T, Leffers T. Self-consistent modelling of the plastic deformation of fcc polycrystals and its implications for diffraction measurements of internal stresses. Acta Mater 1998;46(9):3087–98. [33] Kocks UF, Canova GR, Jonas JJ. Yield vectors in fcc crystals. Acta Metall. 1983;31(8):1243–52.